Beyoncé’s discography. Analysis about how Beyoncé’s music has changed over time.

Beyoncé’s album covers in chronological order

Nowadays, Beyoncé is one of my favorite artists. One of the reasons I like to listen to Beyoncé’s music is because she has such a diverse style in her albums. You can find various music genres come back in her music, like R&B, soul, (dance)pop, hiphop, and funk. So far she has produced the following albums, among others:

• Dangerously In Love (2003)
• B-Day (2006)
• I Am Sacha Fierce (2008)
• 4 (2011)
• Beyoncé (2013)
• Lemonade (2016)
• The Lion King (2019)

Because I like Beyoncé’s diversity in styles and genres between and in her albums, I would like to find out with the SpotifyR API features more about Beyoncé’s music and how her music has changed over time. I will do this by analyzing:

• Spotify features of the seven albums;
• the pitch and timbre features of Beyoncé’s well-known song Love On Top (this song is well knwon for having “definitely the greatest key changes in all pop music”);
• Tempo and timbre features of the albums and songs;

I personally notice a number of differences between her albums, for example that her newer albums contain less acoustic songs and that the overall vibe is less positive. Therefore, I’m going to research the acousticness and valence of her albums. I’m expecting them to have shifted downwards after the release of her album Beyoncé, because this album is known for having a different style compared to Beyoncé’s earlier albums. It’s been said a lot that her later albums contained a different, lower end hip hop bass, so I’m curious if this can somehow be seen in this research. In addition, I’m interested in the danceability, energy, and speechiness of her albums. However, I think these features are very high for all her albums, and that no major differences will be seen.

Albums show a high diversity in character. A shift in valence and acousticness over time.

In these graphs, energy, valence, acousticness, and danceability are shown of Beyoncé’s albums.

• Overall, the scatterplots show that the the songs of every album show a variety in values of energy, valence, acousticness and danceability. This shows that her albums are not static, but that the the albums have a high diversity in character.
• Furthermore, the overall valence level of her albums has shifted downwards after the release of Beyoncé. Before Beyoncé, valence values of her songs were never even close to zero, whereas her last three albums contain more lower-valued valence songs.
• Besides, acousticness has increased a bit in her last two albums, with values above 0.9. This was only seen once in her album Dangerously In Love, but not again untill Lemonade and The Lion King.
• Two songs, Hip Hop Star and Heaven, have been labelled because they’ll come back in a later graph.

No clear changes in speechiness, so this feature has been constant in Beyoncé’s music.

In these graphs, the speechiness is shown of Beyoncé’s albums. Unfortunately, ggplotly removes the y-axis label, but the label speechiness should be there.

Overall, the speechiness of Beyoncé’s albums is more on the lower side, showing a lot of songs with speechiness-values below 0.1. Songs become non-speech-like with values below 0.33. However, her first two albums seem to have more songs with higher speechiness-values (at least with speechiness-values above 0.2). Still, according to Spotify API, this means that her albums contain mostly non-speech-like songs (songs < 0.33). But since her songs have become more hip hop over the years, I would’ve expected a more clear decreasing shift between the albums.

Albums show a diversity in keys, which could mean a higher contrast between the songs in her albums.

In this graph, the musical key percentage of Beyoncé’s albums is shown.

After I created this graph, purely out of interest in the keys she uses for her songs, I was not sure if this graph would be of any use for my research. However, the graph does show something I feel is important for my research, namely the diversity in keys she uses in her albums. Besides the fact that the D#/Eb key is neglected in her songs, all the keys are used in songs of at least three of her albums. I’m not sure if this has in any way effect on my opinion that her albums are so diverse, but it shows some contrast between the songs in her albums.

Structure of Love On Top. Timbre and pitch features shown in self-similarity matrix.

As I already explained in the introduction, I’m interested in the key changes of the song Love On Top, because this song is very well-known for it. Therefore, I will take a look at the chroma and timbre features.In the two self-similarity matrices (SSMs) you can see chroma as well as timbre features of the song Love On Top. Left, you see the chroma features, which show the harmonic and melodic characteristics. On the right, the timbre features are shown, which show the changes in timbre and instrumentation.

• I’ve divided the song into an intro (I), 2x verse (V) and chorus (C), and 4x the first part of the chorus in the end (4xC).
• In the chroma SSM, the dark diagonal lines from 0 to 180 seconds show repetition in pitches. The diagonal lines disappear after approximately 180 seconds. I think this happens because in the last two minutes of the song, the chorus is repeated and goes through four key changes (hence, the 4xC) I’ll explain this in more detail with a chromagram on the next storyboard.
• In the timbre SSM, you mostly see a similar pattern, since the song mostly consists of the chorus (2x a complete chorus, 4x part of the chorus). The intro and verses have a little diversion in color, but that’s not so weird since there is different pattern of vocals and instrumentation.

Chromagram of Love On Top. Here, the modulations are shown more clearly.

In this chromagram, the key changes/modulations of Love On Top (which I explained in the self-similarity matrix) are shown more clearly. As you can see, after 180 seconds the song switches from C major to Db major, D major, Eb major, and lastly to E major.

Songs of the album Beyoncé show a similar tempo, which is a different pattern than the songs of the other albums, which have a more varied tempo.

In this graph, I’ve compared the variables tempo, loudness and duration of two older albums, Dangerously In Love and B-Day Deluxe Edition, with two newer albums, Beyoncé and Lemonade.

The album Beyoncé seems to be different in tempo than the other albums. It’s kind of funny to see that Beyoncé has a lot of songs (seven of the fourteen songs) around a mean tempo of 140 bpm, whereas the songs of the other albums are more spread out.

The constant tempo of Hip Hop Star, and…

As we’ve seen in the graph of the preceding page, the album Beyoncé seems to have a different pattern in tempo compared to the albums Dangerously In Love, B-Day Deluxe Edition, and Lemonade. Therefore, I’d like to compare the tempograms of a song from Beyoncé with a song from one of the other three albums.

• I’ve chosen to compare Hip Hop Star from Dangerously In Love with Heaven from Beyoncé (see next storyboard), since Hip Hop Star has extremely high valence and energy and a low acousticness and Heaven an extremely low valence and energy and a high acousticness (they have the highest and lowest valence in this corpus, which can be seen in the features scatterplot from a few storyboards back).
• As can be seen, the tempo of Hip Hop Star is constant around 82 bpm. This song has a very stable drumbeat throughout the song with no diversions.

the ‘inconsistent’ tempo of Heaven.

Heaven has a more ’noisy’tempo. I think this has to do with the acousticness. Since Heaven is almost 100% acoustic and doesn’t have such a stable beat as Hip Hop Star, it could be harder to determine the exact tempo.

Albums show major difference in timbre coefficient c02.

In this graph, the difference in timbre coefficients between four of Beyoncé’s albums can be seen.

• I know this graph is a bit intense to take in, and I initially wanted to take out two albums because it’s so crowded, but the situation at x = c02 is too interesting to leave out some albums. You can see major differences in the co2 coefficient between the four albums. The c02 describes lower frequencies, which means that in the later albums, there is more emphasized on the lower end (hip hop) bass. The hip hop 808 bass harmonics and tone is very important in Beyoncé’s music production and mixing, so apparently the use of the lower end bass has increased over the years.
• Besides c03, the violin plots of the coefficients are very compact and constant between the four albums, so not significant changes over time.

Ranking Beyoncé’s songs on feature importance. Timbre coefficient c02 stands out.

In this graph, you can see that timbre coefficient c02 stands out as the most important feature. This is remarkable, because c02 has shown to be the most different timbre coefficient between the albums, as you could have seen on the preceding storyboard about timbre coefficients. Maybe this also indicates why this feature is so important in Beyoncé’s music.

Predictability model of Beyoné’s albums with timbre coefficient c02. Algorithm can make a fair prediction on Beyoncé’s older and newer music based on this coefficient.

Here, Beyoncé’s albums and the accompanying songs are divided in two groups, namely old (her first four albums, untill Beyoncé) and new (her last three albums). The algorithm has made predictions about the distribution of the songs in the two groups based on the feature c02 (since this feature stood out in the graphs of the preceding storyboards).

• As can be seen, the predictions are relatively neat. Most of the songs from both “old” and “new” are well predicted, which could means that both groups have a certain recurring c02 which can be recognized by the algorithm.
• So with the last three graphs, We have seen that c02 has decreased drastically over time, that it is the most important feature of Beyoncé’s songs, and that a lot of her songs can be classified into older and newer songs.
• The precision and recall are both very high as you can see in the table below, which makes the predictions more reliable.

# A tibble: 2 x 3
  class precision recall
  <fct>     <dbl>  <dbl>
1 New       0.571  0.64 
2 Old       0.727  0.667

And another predictability model of Beyoncé’s albums, but then with all the features included to see the contrast with the previous predictability model.

There’s only a small difference between the predictions of the older songs, so this means that timbre coefficient c02 is a big factor in the predictions. Precision and recall are a bit higher compared to the previous model.

# A tibble: 2 x 3
  class precision recall
  <fct>     <dbl>  <dbl>
1 New       0.685  0.74 
2 Old       0.809  0.764

Final conclusions from the analysis of Beyoncé’s albums about general features, the song Love On Top, tempo, and timbre features.

In this corpus, I’ve analysed four different things, namely: general features of the albums, pitch and timbre features of the song Love On Top, tempo features of Beyoncé’s albums and songs, and timbre features of Beyoncé’s albums. It was very interesting to see the possibilities of Spotify API together with RStudio. So I’ve not only learned a lot about Beyoncé’s music during this course, but also about programming and visualization.

The main things I’ve learned about Beyoncé’s music is that:

• Valence and acousticness have decreased over time, and energy has been constant;
• Speechiness is very low and has decreased over time;
• Beyoncé doesn’t stick to a few keys, but she varies a lot between her albums and songs;
• And not only between songs, but also within songs: Love On Top modulates no less than 4 times, which makes this song very unique;
• The songs in the album Beyoncé show a lot of similarity in tempo, compared to the variety in tempo in other albums;
• Timbre feature c02 stands out in Beyoncé’s music and has decreased in value over time, which means lower frequencies like bass have become more prominent and that this can define Beyoncé’s music to a high degree (specifically the 808 bass).

It’s impressing how much I’ve learned about Beyoncé’s music only in a few weeks time by Spotify API, and therefore I’m very excited and curious what the future will bring, because I’m sure I’ll use this again for interesting and new musical researches.